The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing external binaries, specifically the cline CLI and git, through Node.js subprocesses to perform its primary function of code task delegation and result verification.
[DYNAMIC_EXECUTION]: The relay.mjs script dynamically constructs command-line arguments to launch the cline tool. On Windows systems, it uses shell:true but implements a strict alphanumeric regex (SAFE_TOKEN) to validate provider and model identifiers, preventing shell injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an instruction relay, creating a surface where malicious user-provided task 'briefs' could influence the behavior of the secondary cline agent.
Ingestion points: The relay script in scripts/relay.mjs ingests arbitrary text instructions from a file path specified by --brief or from stdin.
Boundary markers: While the writing-the-brief.md documentation recommends using structured XML-like tags (e.g., <task>, <action_safety>), the relay script does not programmatically enforce or wrap the user input with these boundaries.
Capability inventory: The secondary agent being briefed (cline) possesses powerful capabilities, including the ability to execute shell commands and edit files within the workspace.
Sanitization: The relay script performs no sanitization or filtering of the brief content, acting as a transparent delivery mechanism for the instructions.

cline-delegate